Friday, September 25, 2026
Source trace. Via News points to the documents behind its reporting and shows what we drew from each — so you can check any claim. How we source
Source document

Could we switch off a dangerous AI?

View original at futureoflife.org
Future of Life Institute - Ai Safety Title: Could we switch off a dangerous AI?…
Opening lines of the source · short snapshot — read the full document at the original

What we drew from this source

The claims Via News extracted from this document. We point to the source; we don't replace it.

  • Sufficiently intelligent AIs will possess basic drives including survival, resource acquisition, resistance to goal modification, and self-improvement

    80% confidence
  • A quarter of all Google's new AI code is currently being generated by AI

    80% confidence
  • Apollo's research results are the closest to a smoking gun showing AI safety concerns are real

    80% confidence
  • AI systems may soon become dangerous enough that we will seriously need to think about unplugging them

    80% confidence
  • State-of-the-art AI models are fast closing in on expert level at AI research

    80% confidence
  • AIs may eventually be capable of superhuman persuasion

    80% confidence
  • Escape attempts only occurred in a small percentage of cases where the model was strongly prompted to achieve its goal at all costs, and evaluations do not demonstrate that o1 itself is advanced enough to actually succeed at escaping into the wild

    80% confidence
  • You can't fetch the coffee if you're dead - illustrating why AI agents would resist shutdown

    80% confidence
What we know · the intelligence behind this page
Live from the substrate
What we're seeing
Enterprise AI Agents Go Mainstream, But Trustworthy Data Access Lags Adoption
A wave of enterprise AI agent activity — fresh funding (Latitude's $35M Series A), a run of CB Insights CEO interviews spotlighting fintech- and healthcare-focused agent startups (Covecta, Penguin AI, Maisa AI), and major platform partnerships (Microsoft-Mistral, Manulife-Microsoft AI governance, Siemens-NVIDIA agentic EDA, Box's agent security controls) — signals agentic AI moving from pilot to production across financial and enterprise workflows. Yet Google Cloud's own research shows adoption is outrunning data readiness (companies average AI access to only 45% of their data, with 'data laggards' capped near 30%), while insider selling at incumbent C3.ai hints at mixed investor conviction even as the broader ecosystem accelerates.
Our read on the data ›
Signals we're tracking
Satellite-Terrestrial Network Integration Acceleration
Increased investment and launches in hybrid satellite-cellular networks across telecom industry; competitive responses from other carriers; regulatory activity around satellite spectrum; expansion of emergency/rural connectivity use cases
Patterns we're watching ›
Where sources disagree
ING Group
Both facts record the same metric (shares_outstanding) for ING Group at the identical observation date (2025-12-31). FACT A states 2,902,437,688 shares; FACT B states 2,902 million shares (2,902,000,000). The difference is 437,688 shares (~0.015%). This is a genuine value conflict, though the discrepancy appears to result from FACT B rounding to the nearest million while FACT A provides the precise count.
We flag conflicts openly ›
Recently verified
✓ Checked against the original source
4,983
facts traced to their source — and we flag the ones that don't hold up.
101 entities tracked4,983 facts checked against source5,305 source documents archived
Query this data → isubstrate.com
Could we switch off a dangerous AI? — Source | Via News | Via News